The transition toward sustainable transportation has accelerated the development of fuel cell hybrid electric vehicles (FCHEVs) where efficient power management plays a crucial role. A critical challenge in these systems lies in dynamically regulating power flow between the fuel cell, battery, and load under varying operating conditions. Traditional control methods like rule-based or PID lack adaptability, causing increased voltage ripple, slow transient response, and suboptimal mode selection in bidirectional DC-DC converters. This paper addresses these challenges by introducing a machine learning-based control strategy that enhances the converter's adaptability and performance. The proposed approach significantly reduces voltage ripple and ensures dynamic mode selection based on power demand, effectively overcoming limitations of conventional control methods. The results confirm the method's suitability for modern FCHEV applications, offering improved efficiency and operational stability.